Prediction Model for Nonlinear Deformation Time Series Based on the Hilbert-Huang Transform

In this paper, we applied the Hilbert–Huang transform method to improve the accuracy of nonlinear deformation predictions. We propose a nonlinear model for prediction based on the multi-scale characteristics of a signal, and used the empirical mode decomposition (EMD) method to decompose the signal. We first applied our method to a simulation of the Lorenz system. Our results show that the EMDs have smaller largest Lyapunov indices than the original signal. We can use this to determine the maximum prediction time for a nonlinear signal. We then constructed a new model based on EMD signals. The results of our experiment demonstrated that this prediction accuracy is perfect. Finally, we used the characteristics of the EMD signals to build the EMD-LLSVM prediction model. Our results show that this model is more accurate than traditional models.